I question Steven Pinker's optimistic prediction of a continuing decline in interpersonal violence, which is based on supposed correlations of mortality data with improved reasoning ability and IQ (Nature 478, 309–311; 2011).

Even if we disregard possible biases in the mortality data and the problems of quantifying reason and IQ over long time spans or across cultures, a sudden shift in innate individual attributes since the mid-twentieth century is unlikely. Pinker's contention that “modern societies have been getting smarter” makes a further dubious leap from an individual to a social intelligence.

More-concrete social features should be considered. The societies in which the decline of interpersonal violence is most marked have become more organized and affluent, with less contention over resources, and violence has mostly been deferred to the state. We know that this deferral can lead to violent events that are more erratic and severe — and even ideological and irrational.

Primitive terrorism and revenge can be blamed for tens of thousands of casualties in Iraq and Afghanistan. We have had the nuclear technology of violence since the Second World War. Occasions on which nuclear weapons were actually used (with some 200,000 civilian casualties), or were close to being used (in the Korean and Vietnam wars and in Cuba), indicate that the statistics for mortality by violence are unstable owing to the escalation and sensitivity of trigger points.

Rather than an irrational optimism, we should adopt a rational, risk-assessment approach that better recognizes the instability of organized and powerful social violence.